Corner sharpening with modified harris corner detection to localize eyes in facial images

摘要

Harris corner detection algorithm has been widely used to find corners in images; it has been used for eye localization in combination with filters. In this paper a new strategy based on filters combination approach is suggested for eyes localization in which filters are used to find and highlight corners of region with local maximum intensity referred here as sharp region. The filtering process proposed here utilizes Harris corner detector in combination with Homomorphic and Tophat-Bothat morphologic filtering. The use of Tophat-Bothat filter enhances image at the first stage resulting in a highlighted region with maximal differences in intensity level after which application of Homomorphic filter extracts the desired region and suppresses others. The proposed method was tested in different applications including biometrics and industrial (mechanical) image analysis in which we used sharp region approach for corner detection. Experimental results, including application on facial image analysis on both color and grayscale, all indicate high performance can be achieved using the proposed method.